recall is live · free · last verified 5 Nov 2026 · viability monitor
recall
Fully-local project memory for Claude Code and Opencode — a resume-ready context.md with zero model tokens spent.
Recall's sharpest decision is what it refuses to do: no LLM call, no API key, no meter. TF-IDF plus TextRank turns a session into a compact context.md, and transcripts stay on disk while you keep paying only the Claude Code subscription. The catch is that extractive summarization can't infer that a renamed module is the same module — treat it as a logbook, not a brain.
Verdict generatedVerified 8d ago · liveness 66/100 · cite: rightaichoice.com/tools/recall
- Developers running Claude Code or Opencode locally on a subscription
- Solo developers burning tokens re-explaining the project at every cold start
- Privacy-conscious coders who won't send transcripts and file paths to a cloud service
- Small teams on medium-to-large local codebases who want a readable, diffable session log
- Teams needing memory synced across devices or shared between people
- Projects where plaintext session transcripts on disk are a security concern
- Anyone needing semantic, LLM-grade recall rather than extractive summarization
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Recall if you need a memory layer that reasons about your codebase or syncs across machines and teammates — its TF-IDF/TextRank summarizer is extractive and its .recall/ store is local-only.
Recall itself is $0 under the MIT license; the real line item is the Claude Code subscription or API credits it runs on. It's aimed at solo devs and small teams who want to stretch that existing spend rather than add a second memory-service bill. Teams needing shared, hosted memory are paying for a different product category entirely.
In short
recall — Fully-local project memory for Claude Code and Opencode — a resume-ready context.md with zero model tokens spent. Best for Developers running Claude Code or Opencode locally on a subscription, Solo developers burning tokens re-explaining the project at every cold start, Privacy-conscious coders who won't send transcripts and file paths to a cloud service. Free to use.
What is recall?
recall — Fully-local project memory for Claude Code and Opencode — a resume-ready context.md with zero model tokens spent. Best for Developers running Claude Code or Opencode locally on a subscription, Solo developers burning tokens re-explaining the project at every cold start, Privacy-conscious coders who won't send transcripts and file paths to a cloud service. Free to use.
Recall is an MIT-licensed plugin that gives Claude Code and Opencode a durable, local record of what each session actually did. It targets developers running Claude Code locally on a subscription who are tired of re-explaining their project every time the assistant opens cold, and who don't want a metered summarizer on the bill. Everything lands in a .recall/ folder inside your project. history.md is append-only — prompts, Claude's replies, files touched and commands run, captured incrementally by the Stop and SessionEnd hooks as you work. context.md is the digest: goal, summary, next steps and open threads, files touched, and where you left off. The summarizer is a classical Python script, not an LLM call. scripts/summarizer.py builds TF-IDF sentence vectors, a cosine-similarity graph between sentences, and runs TextRank power iteration over that graph to score and keep the most central sentences in original order. The result is a resume digest of roughly 1–2K tokens, so picking up costs far less than replaying a full transcript. Setup is deliberately thin: no pip install, no local model to run, no key to configure, works offline on Linux, macOS and Windows. A SessionStart hook surfaces context.md and asks whether to resume and keep logging; /recall:save checkpoints manually before risky work, and auto_save_context: "on_end" regenerates the digest at every session end. It's complementary to CLAUDE.md (hand-written instructions) and --continue (full-transcript replay), not a replacement for either.
Key features
- Fully-local project memory plugin for Claude Code and Opencode
- Automatic session capture via Stop and SessionEnd hooks
- Append-only history.md log of prompts, replies, files touched and commands run
- Local extractive summarization with TF-IDF plus TextRank in scripts/summarizer.py
- Generates a compact context.md resume digest of roughly 1–2K tokens
- Incremental capture — only new turns are appended each time
- SessionStart hook surfaces context.md and asks whether to resume and keep logging
- /recall:save command for manual checkpoints before risky work
- Auto-save at session end via auto_save_context: "on_end"
- Zero model tokens spent capturing or updating memory
- 100% offline — no API key, no external model, nothing sent anywhere
- Plaintext, diffable logs stored in the .recall/ directory
- Runs on Linux, macOS and Windows with no pip install required
- MIT license with published PRIVACY.md, SECURITY.md and CONTRIBUTING.md
Behind the verdict
Pick Recall if the tax you're paying is cold starts. Every time Claude Code reopens a project and you spend the first three prompts restating the goal, architecture and where you stopped, Recall is aimed straight at that. It's also a good fit if you're on a subscription and want to stretch usage limits without adding a second AI bill: capture and summarization run locally, so the memory costs zero model tokens. Pass if you need recall that understands semantics. Recall keeps the most central sentences from a session — it does not reason across your codebase, and it will not connect that the module you renamed last Tuesday is the thing you're asking about today. For that you want a different category of tool entirely. The closest neighbour is the built-in stuff. CLAUDE.md is instructions you curate by hand and Claude follows; --continue replays the entire prior conversation at full token cost. Recall sits between them as the automatic, deterministic record of what happened, condensed to roughly 1–2K tokens. In practice all three coexist fine. Two practical caveats. First, history.md is plaintext on disk — code, paths and occasionally secrets land there, so treat .recall/ like any other sensitive artifact and check PRIVACY.md and SECURITY.md before you commit it. Second, this is a solo/small-team tool with no sync: no shared memory across machines or people, and no vendor SLA behind it. That's the trade for free and local. We'd reach for it on medium-to-large local codebases where sessions are long and context is expensive. We'd skip it on throwaway scripts, and we'd skip it if your assistant isn't Claude Code or Opencode — the hooks are the whole mechanism.
Limitations
- Recall is a deterministic, offline summariser (TF-IDF plus TextRank) rather than a model-driven memory service, so it extracts central sentences rather than reasoning about your code.
- Memory lives per machine and per project in a .recall/ directory with no cloud sync or multi-user sharing.
- Because history.md and context.md are plaintext on disk, anyone with filesystem access can read your prompts, paths and any secrets that appeared in them.
- It only works with Claude Code and Opencode.
as of 2026-10-22
Researching recall? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is recall still active?
recall is live · free · last verified 5 Nov 2026 · viability monitor.
Viability score
How well maintained and how widely used is recall? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: November 2026
How we score →What do users say about recall?
What people actually say about recall — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
Researched26 mentions across 6 sources (Hacker News, Product Hunt, App Store, Stack Overflow, GitHub, Lemmy), 58 more we could not attribute · researched Oct 16, 2026.
Weighted by the 84 posts each of 6 sources contributed.
- +Zero model tokens spent capturing or updating memory — no burn on your subscription
- +100% offline with no API key and nothing sent anywhere, which privacy-conscious devs will love
- +Plaintext, diffable logs in .recall/ (history.md and context.md) are easy to inspect and version
- +MIT-licensed and free, with a PRIVACY.md, SECURITY.md and CHANGELOG.md already in the repo
- +Cross-platform across Linux, macOS and Windows with no pip install or local model to run
- −Community evidence is dominated by coincidental keyword matches, not real user discussion
- −Tiny repo (~28 commits, 44 forks) means limited battle-testing at scale
- −Extractive TF-IDF/TextRank summarization can miss nuance compared to an LLM digest
- −Hook-based capture can silently skip turns if a session ends abnormally
- −No dedicated support channel, SLA, or paid tier — maintainer responsiveness is best-effort
- • No monetary cost, but a real time cost in reading docs and wiring up Stop/SessionEnd hooks
- • If a hook misfires or the repo churns, your 'cost' is debugging your own workflow
How we verified this
Every listed fact about recall — pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 5 Nov 2026. Analysis is drafted with AI models and checked against our editorial standards before publishing; see how we verify.
Verification history
We have re-verified recall 16 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page
- — changed our verdict, summary, description, who it suits, who should skip it, pricing tiers, features
- — changed pricing page
- — changed who should skip it, summary, features, our verdict, description, pricing tiers, who it suits
- — changed pricing page, limitations, models
- — changed pricing tiers, switching out, use cases, description, summary, who should skip it, our verdict, who it suits, setup time, pricing fit, features, hidden costs, switching in
Showing the 6 most recent of 16 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does recall cost in 2026?
recall: Free to use. Pricing last verified 5 Nov 2026.
Plans compared
Every published recall tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Open Source | Free | — | Solo developers and small teams running Claude Code or Opencode locally on a subscription who want durable project memory without another billAdds: Free entry point under the MIT license: full plugin, local TF-IDF/TextRank summarizer, no API key, no usage cap beyond your own disk |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where recall's pricing actually pencils out — and where peers do it cheaper.
Recall itself is $0 under the MIT license; the real line item is the Claude Code subscription or API credits it runs on. It's aimed at solo devs and small teams who want to stretch that existing spend rather than add a second memory-service bill. Teams needing shared, hosted memory are paying for a different product category entirely.
Has recall's pricing changed?
No plan price for recall has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick recall?
Pick recall if you are Developers running Claude Code or Opencode locally on a subscription, Solo developers burning tokens re-explaining the project at every cold start, Privacy-conscious coders who won't send transcripts and file paths to a cloud service, Small teams on medium-to-large local codebases who want a readable, diffable session log.
Look elsewhere if you are Teams needing memory synced across devices or shared between people, Projects where plaintext session transcripts on disk are a security concern, Anyone needing semantic, LLM-grade recall rather than extractive summarization, Developers using an AI coding assistant other than Claude Code or Opencode.
Skip it if Skip Recall if you need a memory layer that reasons about your codebase or syncs across machines and teammates — its TF-IDF/TextRank summarizer is extractive and its .recall/ store is local-only.
GeneratedReal-world workflow fit
Concrete scenarios for the personas recall actually fits — and what changes day-one when you adopt it.
You finish a session mid-refactor, close the terminal, and open Claude Code the next morning on the same repo.
Outcome: The SessionStart hook surfaces context.md — goal, summary, next steps, files touched — so you resume in one prompt instead of re-explaining the project, and the capture cost zero model tokens.
You want session memory but won't pipe transcripts containing internal paths to a hosted memory API.
Outcome: Recall writes history.md and context.md into .recall/ on disk and never makes a network call, so the log stays readable only to you and is diffable in git.
You're at a clean stopping point and want the next session to know exactly where you stopped.
Outcome: You run /recall:save to write a manual checkpoint, and auto_save_context: "on_end" ensures the session end appends the remaining turns to history.md.
Use cases
- Start a new Claude Code session and resume from previous work without re-explaining the project
- Cut token spend on your Claude Code subscription by loading a ~1–2K-token context.md instead of full re-explanations
- Keep a private, offline log of coding sessions for personal reference
- Add persistent memory to a local dev workflow with zero cloud dependency
- Give Claude Code long-term memory across sessions on a medium-to-large codebase at no extra cost
- Track how project context evolves over time through condensed session summaries
- Checkpoint manually before a risky refactor with /recall:save so the next session knows where you stopped
Setup time & first value
How long it actually takes to get something useful out of recall — broken out by persona, not the marketing-page minute.
Claude Code users: minutes. The plugin loads with no pip install, no local model and no API key, and starts capturing from the first session via the Stop and SessionEnd hooks. Opencode users add a little extra time to wire up the integrations/opencode folder. First useful resume comes at the end of your first session, once context.md has something to load.
Switching to or from recall
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From manual CLAUDE.md notes: keep the rules in CLAUDE.md and let Recall record what actually happened in each session — the README frames them as complementary, not competing
- →From re-explaining your project each session: let the SessionStart hook load context.md instead of retyping the brief
- →From a cloud memory service: copy any context you want to keep into the project, then let history.md and context.md build up locally with no network calls
- ↗To a hosted memory service: your .recall/history.md and context.md are plaintext Markdown, so you can paste or ingest them into the new store
- ↗To CLAUDE.md-only workflow: delete the .recall/ directory and hand-curate the notes you want Claude to follow
Integrations
Resources & guides
- Resourcegithub.com
Recall · recall
Helpful link from github.com
- Resourcegithub.com
README · recall
Helpful link from github.com
- Resourcegithub.com
Claude Code Context Plugin · recall
Helpful link from github.com
- Resourcegithub.com
CONTRIBUTING · recall
Helpful link from github.com
- Resourcegithub.com
CHANGELOG · recall
Helpful link from github.com
Tutorials & Learning
YouTube returned 6 videos for “recall”, and we withheld 6: 6 could not be judged, because “recall” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about recall.
Official links
What are the best recall alternatives?
The closest alternatives to recall we list are Codey, Tabnine and Sourcegraph Cody. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Codey | Freemium | Pricing not published | — |
| Tabnine | Freemium | Pricing not published | — |
| Sourcegraph Cody | Paid | Pricing not published | — |
Alternatives to recall
View all
Codey
Codey brings local-first agentic coding to your machine: 25 agent CLIs and 200+ models on your own API keys for $10 flat.
Tabnine
Enterprise AI coding assistant with a Context Engine for governed, air-gapped code completion
Sourcegraph Cody
Sourcegraph Cody brings whole-codebase context to AI coding chat, completions, and agentic batch changes across hundreds of repos.
Tools that pair well with recall
Common stack mates teams adopt alongside recall, with the specific reason each pairing earns its keep.
Codey
Codey brings local-first agentic coding to your machine: 25 agent CLIs and 200+ models on your own API keys for $10 flat.
Tabnine
Enterprise AI coding assistant with a Context Engine for governed, air-gapped code completion
Sourcegraph Cody
Sourcegraph Cody brings whole-codebase context to AI coding chat, completions, and agentic batch changes across hundreds of repos.
recall vs its closest rival
recall's closest head-to-head is with Bito — These look similar from a distance — both attack the same waste, the agent re-reading and re-learning your project — but they're bought by different people. If you're a solo dev or small team on Claude Code or Opencode who wants zero cost, zero cloud, and plaintext memory files you can diff, take Recall; it's free and does exactly one job locally. If you're an engineering org watching agent spend climb across multiple repos and multiple agents, Bito Governor is the one with routing, budgets, and admin analytics — and it will ask you to index your codebase before it pays off. Don't buy Bito for a single repo; don't expect Recall to give you fleet-wide spend control. Verdict generated
Frequently asked questions
For makers of recall
- Get a Verified Listing for recall ($39): a written decision within 1 working day.
- Free listing for another tool: Free reviews are decided in order, within 3 weeks of the review starting.
Used recall? Help shape our editorial sentiment research.